Social Media Mining Behavior Analytics Examples of Behavior Analytics • What motivates individuals to join an online group? • When individuals abandon social media sites, where do they migrate to? • Can we predict box office revenues for movies from tweets posted by individuals? Social Media Mining Measures Behavior and Analytics Metrics 22 • To answers these questions we need to analyze or predict behaviors on social media. • Individuals exhibit different behaviors in social media: as individuals • or as part of a broader collective behavior. • When discussing individual behavior, our focus is on one individual. • Collective behavior emerges when a population of individuals behave in a similar way with or without coordination or planning. Social Media Mining Measures Behavior and Analytics Metrics 33 Our Goal To analyze, model, and predict individual and collective behavior Social Media Mining Measures Behavior and Analytics Metrics 44 Individual Behavior Social Media Mining Measures Behavior and Analytics Metrics 55 Types of Individual Behavior • User-User (link generation) befriending, sending a message, playing games, following, inviting • User-Community: joining or leaving a community, participating in community discussions • User-Entity (content generation) writing a post, posting a photo Social Media Mining Measures Behavior and Analytics Metrics 66 I. Individual Behavior Analysis Social Media Mining Measures Behavior and Analytics Metrics 77 Community Membership in Social Media • Why do users join communities? – What factors affect the community-joining behavior of individuals? • We can observe users who join communities and determine the factors that are common among them • We require a population of users, a community C, and community membership information (i.e., users who are members of C). – Communities can be implicit: One can think of individuals buying a product as a community, and people buying the product for the first time as individuals joining the community. – To distinguish between users who have already joined the community and those who are now joining it, we need community memberships at two different times t1 and t2, with t2> t1. – At t2, we determine users such as u who are currently members of the community, but were not members at t1. These new users form the subpopulation that is analyzed for community-joining behavior. Social Media Mining Measures Behavior and Analytics Metrics 88 Community Membership in Social Media • Hypothesis: individuals are inclined toward an activity when their friends are engaged in the same activity. • A factor that plays a role in users joining a community is the number of their friends who are already members of the community. • In data mining terms, this translates to using the number of friends of an individual in a community as a feature to predict whether the individual joins the community (i.e., class attribute). Social Media Mining Number of Friends vs Probability of Joining a Community Measures Behavior and Analytics Metrics 99 Even More Features Social Media Mining Measures Behavior and Analytics Metrics 10 10 Feature Importance Analysis • which feature can help best determine whether individuals will join or not? • We can use any feature selection algorithm. Feature selection algorithms OR • We can use a classification algorithm, such as decision tree learning – Most important Features are Ranked Higher Social Media Mining Measures Behavior and Analytics Metrics 11 11 Decision Tree for Joining a Community • Are these features well-designed? – We can evaluate using classification performance metrics. Social Media Mining Measures Behavior and Analytics Metrics 12 12 Behavior Analysis Methodology • An observable behavior: the behavior needs to be observable. E.g., accurately observing the joining of individuals (and possibly their joining times). • Features: design relevant data features (covariates) that may or may not affect (or be affected by) the behavior – [we need a domain expert for this step] • Feature-Behavior Association: Find the relationship between features and behavior. E.g., use decision Tree • Evaluation: the findings are due to the features and not to externalities. E.g., use classification accuracy – We can also use randomization tests – Or causality testing algorithms Social Media Mining Measures Behavior and Analytics Metrics 13 13 Granger Causality • Consider a linear regression model – We can predict Yt+1 by using either Y1, Y2 … Yt or a combination of X1, X2 … Xt and Y1, Y2 … Yt. – If then X Granger Causes Y • (why is this not causality) Social Media Mining Measures Behavior and Analytics Metrics 14 14 II. Individual Behavior Modeling Social Media Mining Measures Behavior and Analytics Metrics 15 15 Individual Behavior Modeling • Models in Economics, Game Theory, and Network Science • We can use: – Threshold Models: We need to learn thresholds and weights – Then, Wij can be defined as the • fraction of times user i buys a product and user j buys the same product soon after that • When is soon? – Similarly, thresholds can be estimated by taking into account the average number of friends who need to buy a product before user i decides to buy it. – What if no friends / or friends don’t buy the same products? Then we can find the most similar individuals or items (similar to collaborative filtering methods) – Cacade Models Social Media Mining Measures Behavior and Analytics Metrics 16 16 III. Individual Behavior Prediction Social Media Mining Measures Behavior and Analytics Metrics 17 17 Individual Behavior Prediction • Most behaviors result in newly formed links in • social media. – It can be a link to a user, as in befriending behavior; – A link to an entity, as in buying behavior; or – A link to a community, as in joining behavior. • We can formulate many of these behaviors as a link prediction problem. Social Media Mining Measures Behavior and Analytics Metrics 18 18 Link Prediction • Most behaviors result in newly formed links in • social media. – It can be a link to a user, as in befriending behavior; – A link to an entity, as in buying behavior; or – A link to a community, as in joining behavior. • We can formulate many of these behaviors as a link prediction problem. Social Media Mining Measures Behavior and Analytics Metrics 19 19 Link Prediction - Setup • Link prediction assumes a graph G(V; E). • Let e(u,v) represent an interaction (edge) between nodes u and v, and let t(e) denote the time of the interaction. • Let G[t1,t2] represent the subgraph of G such that all edges are created between t1 and t2 – i.e., for all edges e in this subgraph, t1 < t(e) < t2). • Now given four time stamps t11 < t12 < t21 < t22 a link prediction algorithm is given the subgraph G[t11,t12] (training interval) and is expected to predict edges in G[t21,t22] (testing interval). • We can only predict edges for nodes that are present during the training period • Let G(Vtrain; Etrain) be our training graph. Then, a link prediction algorithm generates a sorted list of most probable edges in Social Media Mining Measures Behavior and Analytics Metrics 20 20 Link Prediction - Algorithms • Assign to every edge e(x,y) • Edges sorted by this value in decreasing order will create our ranked list of predictions • Note that any similarity measure between two nodes can be used for link prediction; therefore, methods discussed in Chapter 3 are of practical use here. • We will review some well-known methods – Node Neighborhood-Based Methods – Path-Based Methods Social Media Mining Measures Behavior and Analytics Metrics 21 21 Computing Link Scores • Common Neighbors: the more common neighbors that two nodes share, the more similar they are • Jaccard Similarity: the likelihood of a node that is a neighbor of either x or y to be a common neighbor. Social Media Mining Measures Behavior and Analytics Metrics 22 22 More Measures • Adamic-Adar:if two individuals share a neighbor and that neighbor is a rare neighbor, it should have a higher impact on their similarity. • Preferential Attachment: nodes of higher degree have a higher chance of getting connected to incoming nodes Social Media Mining Measures Behavior and Analytics Metrics 23 23 Example • Compute the edge score for edge (5,7) Social Media Mining Measures Behavior and Analytics Metrics 24 24 Path-Based Measures • Katz Measure: • denotes the number of paths of length l between x and y. is a constant that exponentially damps longer paths. – When small the measure reduces to common neighbor measure – It can be reformulated in Closed Form as Social Media Mining Measures Behavior and Analytics Metrics 25 25 Path-Based Measures • Hitting Time and Commute Time: Consider a random walk that starts at node x and moves to adjacent nodes uniformly. Hitting time is the expected number of random walk steps needed to reach y starting from x. – a smaller hitting time implies a higher similarity; therefore, a negation can turn it into a similarity measure – If y is highly connected random walks are more likely to visit y • We can normalize it using the stationary probability Social Media Mining Measures Behavior and Analytics Metrics 26 26 • Hitting time is not symmetric, we can use commute time instead or its normalized version • Some other examples – Rooted PageRank: the stationary probability of y, when at each random walk run you can jump to x with probability P and to a random node with probability 1-P – SimRank: Recursive definition of similarity Social Media Mining Measures Behavior and Analytics Metrics 27 27 Link Prediction - Evaluation • After one of the aforementioned measures is selected, a list of the top most similar pairs of nodes are selected. • These pairs of nodes denote edges predicted to be the most likely to soon appear in the network. • Performance (precision, recall, or accuracy) can be evaluated using the testing graph and by comparing the number of the testing graph’s edges that the link prediction algorithm successfully reveals. • Performance is usually very low, since many edges are created due to reasons not solely available in a social network graph. • So, a common baseline is to compare the performance with random edge predictors and report the factor improvements over random prediction. Social Media Mining Measures Behavior and Analytics Metrics 28 28 Collective Behavior Social Media Mining Measures Behavior and Analytics Metrics 29 29 Collective Behavior • First Defined by sociologist Robert Park • Collective Behavior: A group of individuals behaving in a similar way • It can be planned and coordinated, but often is spontaneous and unplanned • Examples: • Individuals standing in line for a new product release • Posting messages online to support a cause or to show support for an individual Social Media Mining Measures Behavior and Analytics Metrics 30 30 I. Collective Behavior Analysis Social Media Mining Measures Behavior and Analytics Metrics 31 31 Collective Behavior Analysis • We can analyze collective behavior by analyzing individuals performing the behavior • We can then put together the results of these analyses • The result would be the expected behavior for a large population • OR, we can analyze the population as a whole • Not very popular for analysis, as individuals are ignored • Popular for Prediction purposes Social Media Mining Measures Behavior and Analytics Metrics 32 32 Collective Behavior Analysis • We can analyze collective behavior by analyzing individuals performing the behavior • We can then put together the results of these analyses • The result would be the expected behavior for a large population • OR, we can analyze the population as a whole • Not very popular for analysis, as individuals are ignored • Popular for Prediction purposes Social Media Mining Measures Behavior and Analytics Metrics 33 33 Collective Behavior Analysis - Example • Users migrate in social media due to their limited time and resources • Sites are often interested in keeping their users, because they are valuable assets that help contribute to their growth and generate revenue by increasing traffic • Two types of migrations: • Site migration: For any user who is a member of two sites s1 and s2 at time ti, and is only a member of s2 at time tj > ti, then the user is said to have migrated from site s1 to site s2. • Attention Migration: For any user who is a member of two sites s1 and s2 and is active at both at time ti, if the user becomes inactive on s1 and remains active on s2 at time tj > ti, then the user’s attention is said to have migrated away from site s1 and toward site s2. Social Media Mining Measures Behavior and Analytics Metrics 34 34 Collective Behavior Analysis - Example • Activity (or inactivity) of a user can be determined by observing the user’s actions performed on the site. • We can consider a user active in interval [t, t+X ], if the user has performed at least one action on the site during this interval. Otherwise, the user is considered inactive. • The interval could be measured at different granularity, such as days, • weeks, months, and years. – It is common to set = 1 month. • We can analyze migrations of individuals and then measure the rate at which the population of these individuals is migrating across sites. – We can use the methodology for individual behavior analysis Social Media Mining Measures Behavior and Analytics Metrics 35 35 The Observable Behavior • Sites migrating is rarely observed, while attention migration is clearly observable. • We need to take multiple steps to observe it: • Users are required to be identified on multiple networks • Huan.liu1 on Facebook is Liuhuan on Twitter Social Media Mining Measures Behavior and Analytics Metrics 36 36 Features • User Activity: more active users are less likely to migrate • e.g., number of tweets, posts, photos • User Network Size: a user with more social ties (i.e., friends) in a social network is less likely to move • e.g., number of friends • User Rank: A user with high status in a network is less likely to move to a new one where he or she must spend more time getting established. • e.g., centrality scores • External rank: your citations, how many have referred to your article, ... Social Media Mining Measures Behavior and Analytics Metrics 37 37 Feature-Behavior Association • Given two snapshots of a network, we know if users migrated or not. • Let vector indicate whether any of our n users have migrated or not. • Let be the features collected (activity, friends, rank) for any one of these users at time stamp t. • The correlation between features X and labels Y can be computed via logistic regression. • How can we verify that this correlation is not random? Social Media Mining Measures Behavior and Analytics Metrics 38 38 Evaluation Strategy • To verify if the correlation between features and the migration behavior is not random • we can construct a random set of migrating users and compute X_Random and Y_Random for them as well • Find the correlation between these random variables (e.g., regression coefficients) and it should be significantly different from what we obtained using real-world observations • We can use Chi-square test for significance testing Social Media Mining Measures Behavior and Analytics Metrics 39 39 II. Collective Behavior Modeling Social Media Mining Measures Behavior and Analytics Metrics 40 40 Collective Behavior Modeling • Collective behavior can be conveniently modeled using some of the techniques discussed in Chapter 4, “Network Models”. • We want models that can mimic characteristics observable in the population. • In network models, node properties rarely play a • role; therefore, they are reasonable for modeling collective behavior. Social Media Mining Measures Behavior and Analytics Metrics 41 41 III. Collective Behavior Prediction Social Media Mining Measures Behavior and Analytics Metrics 42 42 Collective Behavior Prediction • From previous chapters, we could use • Linear Influence Model (LIM) • Epidemic Models • Collective behavior can be analyzed either in terms of individuals performing the collective behavior or based on the population • as a whole. • It is more common to consider the population as a whole • when predicting collective behavior, we are interested in predicting the intensity of a phenomenon, which is due to the collective behavior of the population (e.g., how many of them will vote?) • we can utilize a data mining approach where features that describe the population well are used to predict a response variable (i.e., the intensity of the phenomenon). • A training-testing framework or correlation analysis is used to determine the generalization and the accuracy of the predictions. Social Media Mining Measures Behavior and Analytics Metrics 43 43 Predicting Box Office Revenue for Movies 1. Set the target variable that is being predicted. In this case, it is the revenue that a movie produces. 1. 2. 3. 4. The revenue is the direct result of the collective behavior of going to the theater to watch the movie. Determine the features in the population that may affect the target variable. Predict the target variable using a supervised learning approach, utilizing the features determined in step 2. Measure performance using supervised learning evaluation. • the average hourly number of tweets related to the movie for each of the seven days • prior to the movie opening (seven features) and the number of opening theaters for the movie (one feature). • The predictions using this approach are closer to reality than that of the Hollywood Stock Exchange (HSX), which is the gold standard for predicting revenues for movies Social Media Mining Measures Behavior and Analytics Metrics 44 44 Generalizing the Idea • Target variable y • Some feature A that quantifies the attention • Some feature P that quantifies the publicity • Train a regression model Social Media Mining Measures Behavior and Analytics Metrics 45 45